Manufacturing system energy consumption evaluation method based on Bayesian network
A Bayesian network-based method addresses the challenge of high energy consumption and uncertainty in aluminum extrusion by predicting energy use, enhancing energy management and optimization.
Patent Information
- Application Number
- CN202411444754.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-16
- Publication Date
- 2024-12-20
AI Technical Summary
The aluminum extrusion manufacturing industry faces high energy consumption and uncertainty due to frequent changes in production orders and variable raw materials, making it difficult to plan and optimize energy use effectively.
A method using a Bayesian network to evaluate and predict energy consumption by modeling uncertainties in the manufacturing process, including heating, extrusion, cooling, and transportation stages, and integrating multidimensional uncertainty factors through kernel density estimation and Monte Carlo sampling.
The method provides accurate predictions of energy consumption under uncertain conditions, enabling better energy management and optimization in aluminum extrusion systems.
Smart Images
Figure CN119168159A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of aluminum profile extrusion manufacturing, and in particular to a manufacturing system energy consumption evaluation method based on a Bayesian network. Background Art
[0002] According to the statistics of "Energy Consumption Index of Nonferrous Metal Processing Enterprises" (YS / T109-92), the energy consumption of the extrusion process of major aluminum processing enterprises in China is 1571~5405kg standard coal / t, which is a typical energy-intensive industry. The 3600T extrusion molding workshop of a large aluminum profile enterprise in South China has been in operation since 2012. It is 16.5 meters high and has a construction area of 14,368 square meters. After energy audit, it was found that the electricity consumption of the extrusion manufacturing system accounted for 55.59% of the total electricity consumption of the enterprise, the natural gas consumption accounted for 84.95% of the total natural gas consumption of the enterprise, and the water consumption accounted for 15.86% of the total water consumption of the enterprise. It is the main energy-consuming department of aluminum profile production. In the past few years, with the rapid growth of raw material prices, it has affected almost all production processes, because the aluminum extrusion manufacturing system depends on energy supply (electricity, natural gas) and aluminum ingot raw material supply. Therefore, all production processes are aimed at reducing energy consumption and striving to reduce carbon dioxide emissions into the environment. This paper studies the electrical energy consumption during the production of aluminum extrusion and how the process of extrusion manufacturing systems can be adjusted to achieve higher energy efficiency. Only the electrical energy consumption was evaluated.
[0003] The market competition in the aluminum extrusion industry is very fierce. Aluminum extrusion production plants are reluctant to lose any customer. The time period between ordering extruded aluminum and its production is usually less than 24 hours, and the different needs of customers must be met in a short period of time. This leads to frequent changes in production orders during the extrusion production process, and it is impossible to plan production orders in advance. The aluminum bar raw materials used in the aluminum extrusion manufacturing system are determined by the customer's production order, so some uncertainties in the production process are related to highly variable raw materials, which will have a great impact on the performance of energy-saving processing. Therefore, these uncertainties must be dealt with in the daily production of the extrusion manufacturing system. Summary of the invention
[0004] In view of the shortcomings of the prior art, an embodiment of the present invention aims to provide a manufacturing system energy consumption assessment method based on a Bayesian network to solve the problems in the above-mentioned background technology.
[0005] To achieve the above object, the present invention provides the following technical solutions:
[0006] A method for evaluating energy consumption of a manufacturing system based on a Bayesian network comprises the following steps:
[0007] Step 1: Obtain the aluminum profile extrusion manufacturing process, which includes a heating furnace, an extruder, processing equipment and a conveying device;
[0008] Step 2: Obtain the total energy consumption E in the manufacturing process m , the total energy consumption E m Including processing energy consumption E ope , Standby energy consumption E rea , Pretreatment energy consumption E pre , Transportation energy consumption E tran and public energy consumption E pub ;
[0009] Step 3: Build an evaluation model based on manufacturing process and total energy consumption.
[0010] Step 4: Obtain the sources of uncertainty and build a prediction model based on accidental uncertainty;
[0011] Step 5: construct a process Bayesian network based on the multidimensional prediction model, and generate samples of the posterior distribution of process parameters based on the process Bayesian network;
[0012] Step 6: Use kernel density estimation and Monte Carlo sampling methods to perform Bayesian forward uncertainty propagation, and obtain the posterior distribution of process parameters to estimate the posterior distribution of the total energy consumption of the extrusion molding manufacturing system.
[0013] As a further solution of the present invention, the process of the manufacturing process in step one is: heating of aluminum rods → extrusion → cooling and quenching → stretching and straightening → sawing → inspection → artificial aging → inspection → warehousing.
[0014] As a further solution of the present invention, the processing energy consumption E in step 2 is ope is the energy consumption of the extruder and the three subsequent processes when machining aluminum profile product i in order e,
[0015]
[0016] As a further solution of the present invention, the transport energy consumption E in step 2 is tran The energy consumed when the product is transferred between the various devices in the extrusion manufacturing system.
[0017]
[0018] As a further solution of the present invention, the public energy consumption E in step 2 pub It is the sum of energy consumed by lighting, cooling, water supply and air supply to maintain a constant temperature in the production environment.
[0019]
[0020] As a further solution of the present invention, the Bayesian network technology in step 4 expresses the multivariate joint probability density function as the product of marginal and conditional probability density functions, which is used for information fusion and integration of multidimensional uncertainty factors.
[0021] As a further solution of the present invention, the measurement error of the process parameters in step six is assumed to be normally distributed with a zero mean and a standard deviation of 0.01.
[0022] In summary, compared with the prior art, the embodiments of the present invention have the following beneficial effects:
[0023] An energy consumption uncertainty prediction model was developed based on the Bayesian network method, which can predict the energy consumption of the extrusion manufacturing system with high accuracy under uncertainty conditions.
[0024] In order to more clearly illustrate the structural features and effects of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] Figure 1 This is an energy consumption flow analysis diagram of the manufacturing system of an embodiment of the invention.
[0026] Figure 2 This is an energy consumption uncertainty assessment model for an extrusion manufacturing system in an embodiment of the invention.
[0027] Figure 3 It is a Bayesian network model of the extrusion manufacturing system process in the embodiment of the invention.
[0028] Figure 4 It is the prior distribution and posterior distribution of the total energy consumption of the extrusion molding manufacturing system in the embodiment of the invention. DETAILED DESCRIPTION
[0029] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0030] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.
[0031] In one embodiment, a manufacturing system energy consumption assessment method based on Bayesian network is provided. Figure 1 to Figure 4 , including the following steps:
[0032] Step 1: Obtain the aluminum profile extrusion manufacturing process, which includes a heating furnace, an extruder, processing equipment and a conveying device;
[0033] Step 2: Obtain the total energy consumption E in the manufacturing process m, the total energy consumption E m Including processing energy consumption E ope , Standby energy consumption E rea , Pretreatment energy consumption E pre , Transportation energy consumption E tran and public energy consumption E pub ;
[0034] Step 3: Build an evaluation model based on manufacturing process and total energy consumption.
[0035] Step 4: Obtain the sources of uncertainty and build a prediction model based on accidental uncertainty;
[0036] Step 5: construct a process Bayesian network based on the multidimensional prediction model, and generate samples of the posterior distribution of process parameters based on the process Bayesian network;
[0037] Step 6: Use kernel density estimation and Monte Carlo sampling methods to perform Bayesian forward uncertainty propagation, and obtain the posterior distribution of process parameters to estimate the posterior distribution of the total energy consumption of the extrusion molding manufacturing system.
[0038] For further information, see Figure 1 to Figure 4 The manufacturing process in step one is as follows: heating of aluminum bars → extrusion → cooling and quenching → stretching and straightening → sawing → inspection → artificial aging → inspection → warehousing.
[0039] For further information, see Figure 1 to Figure 4 , the processing energy consumption E in step 2 ope is the energy consumption of the extruder and the three subsequent processes when machining aluminum profile product i in order e,
[0040]
[0041] For further information, see Figure 1 to Figure 4 , the energy consumption of transportation in step 2 is E tran The energy consumed when the product is transferred between the various devices in the extrusion manufacturing system.
[0042]
[0043] For further information, see Figure 1 to Figure 4 , the public energy consumption E in step 2 pub It is the sum of energy consumed by lighting, cooling, water supply and air supply to maintain a constant temperature in the production environment.
[0044]
[0045] For further information, see Figure 1 to Figure 4In the step 4, the Bayesian network technology expresses the multivariate joint probability density function as the product of the marginal and conditional probability density functions, which is used for information fusion and the integration of multidimensional uncertainty factors.
[0046] For further information, see Figure 1 to Figure 4 The measurement error of the process parameters in step six is assumed to be normally distributed with zero mean and a standard deviation of 0.01.
[0047] In this embodiment, a complete aluminum extrusion manufacturing system is usually composed of physical entities such as a heating furnace, an extruder, processing equipment, and a conveying device. Among them, the aluminum bar heating furnace is the starting point of the manufacturing system, providing the manufacturing system with aluminum bars to be processed; the aging furnace is the end of the manufacturing system, and performs artificial aging and stress relief annealing on the aluminum profiles. The production of the aluminum profile extrusion manufacturing system is highly dependent on energy. The various profiles produced can be classified into the same family according to the order, which we represent as the total product i. An energy consumption assessment model for the extrusion manufacturing system is developed, which takes into account the uncertainty of production orders, mold replacement, equipment failure, and environmental factors.
[0048] The process flow of the extrusion molding workshop is: aluminum bar heating → extrusion → cooling and quenching (air cooling, water cooling, mist cooling) → stretching and straightening → sawing → inspection → artificial aging → inspection → storage, such as Figure 1 The figure shows the energy consumption analysis of the equipment involved in the process flow of the extrusion molding workshop. Among them, the environmentally friendly and energy-saving rod furnace burns gas to heat the aluminum rod to 480℃~520℃, and the heat circulation fan, rod transfer mechanism, and motor-driven hydraulic shearing components are the main power-consuming components of the rod furnace; the mold furnace uses electric heating to 460~500℃ and keeps warm for 2~4 hours; the extruder uses a motor-driven hydraulic system to extrude aluminum profiles, and its energy consumption accounts for more than 60% of the total energy consumption of the extrusion molding workshop, and it is the main energy-consuming equipment in the workshop; the automatic quenching system is mainly used for air cooling and mist cooling, and the fan and water pump are its power-consuming components; the aging furnace uses natural gas to heat to 180~200℃, and then uses electric heating and heat preservation to carry out artificial aging of aluminum profiles and stress relief annealing; power-consuming equipment such as air conditioners and exhaust fans are used to maintain the ambient temperature of the manufacturing system at 15~30℃; in addition, there are other auxiliary power-consuming equipment such as automatic double traction machines, induction conveying scaffolds, stretching machines, finished product saws, and bridge cranes. The aluminum profile extrusion manufacturing system is not only energy-intensive but also information-intensive. Its production process, energy consumption flow and factors affecting energy consumption are very complex, which leads to the nonlinearity and randomness of massive energy consumption data.
[0049] The total energy consumption Em of the extrusion manufacturing system can be divided into five parts, corresponding to the equipment such as Figure 1 As shown:
[0050] Processing energy consumption E opeEnergy consumption of the extruder and three subsequent process machines ③④⑤ when processing aluminum profile product i in order e.
[0051]
[0052] Standby energy consumption E rea It is the energy consumption of the extrusion manufacturing system equipment when it is in a non-shutdown state, mainly including the insulation state of three types of heating equipment ①⑥⑦, short-term equipment failure, and the standby state when the extruder replaces the mold.
[0053]
[0054] Pretreatment energy consumption E pre Energy consumption for cleaning and preheating when machine j changes aluminum profile product i according to the needs of different orders e.
[0055]
[0056] Transport energy consumption E tran It is the energy consumed when the product is transferred between the various devices in the extrusion manufacturing system.
[0057]
[0058] Public energy consumption E pub It refers to the energy that must be consumed to maintain the normal operation of the manufacturing system, mainly including the total energy consumed for lighting, cooling, water supply, gas supply, etc. to maintain a constant temperature in the production environment.
[0059]
[0060] The total energy consumption E of the aluminum profile extrusion manufacturing system is:
[0061]
[0062] t is the operating time of the manufacturing system, i is the type of aluminum profile product, is the processing time of product i on machine j, is the energy consumption per unit time when product i is processed on machine j, is the standby time of product i on machine j, is the standby energy consumption per unit time of machine j, Pre-processing time when changing product i for machine j, is the energy consumption per unit time of preprocessing when machine j changes product i, For products from machine M j-1 To M j The transportation time between For products from machine M j-1 To M jUnit mass transport energy consumption, m i For the quality of product i, is the public system power consumption in period t.
[0063] The research process of the energy consumption uncertainty assessment model of aluminum profile extrusion manufacturing system is as follows: Figure 2 As shown in the figure, we first construct a process Bayesian network based on the extrusion process knowledge and the manufacturing system energy consumption model derived in Section 2. The extrusion process knowledge is used to capture the dependencies between variables, and the extrusion process data is used to capture the remaining dependencies. Then, Bayesian reasoning is performed, and the process parameter observation data is used to calibrate the unknown distribution parameters of the process parameters in the model (estimate the posterior distribution of the process parameters). Finally, Bayesian forward uncertainty propagation is performed, using the Monte Carlo sampling method, in order to obtain the posterior distribution of the manufacturing system energy consumption.
[0064] Aleatoric uncertainty can be modeled as random variables of time-independent quantities, random processes of time-varying quantities, and random fields of space-varying quantities. Bayesian methods can be used to model the lack of knowledge about unknown quantities, thereby achieving modeling of epistemic uncertainty. Since Bayesian methods can systematically integrate available information, they can be used to quantify the reduction in uncertainty when more information becomes available. Interval variables and evidence theory can also be used to model variables with higher uncertainty. For modeling of model uncertainty, the difference between model predictions and experimental observations can be quantified using Bayesian calibration.
[0065] Bayesian network technology expresses the multivariate joint probability density function as the product of marginal and conditional probability density functions, which can be used for information fusion and the integration of multidimensional uncertainty factors. At each level of the uncertainty analysis problem, an agent model can be constructed to replace the independent simulation model (which is very computationally intensive). The input and output of the agent model at different levels can become nodes of the Bayesian network and are connected together in a multi-level manner through the Bayesian network. In this case, we are not sure about the connection between different nodes, but we can learn the connection between these nodes through Bayesian network learning technology. The multidimensional sources of uncertainty occurring in different processes of the aluminum extrusion manufacturing system are not combined in a simple way. This combination can be linear, nonlinear, iterative or nested. In order to solve this problem, the present invention adopts a Bayesian network method to aggregate the multidimensional sources of uncertainty in the manufacturing system.
[0066] The extrusion molding manufacturing system is a manufacturing network involving 7 sub-processes. The total energy consumption of the manufacturing system is the sum of the energy consumption of each sub-process and the common energy consumption. Figure 3As shown in the figure, this paper constructs a Bayesian network of the extrusion manufacturing system process to represent the dependency relationship between the variables in the total energy consumption equation (6) of the manufacturing system, and performs uncertainty analysis on the energy consumption of the extrusion molding manufacturing system in a modular and hierarchical manner. According to the extraction results of the energy consumption characteristics of the extrusion molding manufacturing system, the obtained energy consumption data are classified into five categories: determined values, observed data, calibration parameters, accidental uncertainty (distribution parameters are known), and accidental uncertainty (distribution parameters are unknown). The process parameters of accidental uncertainty (distribution parameters are unknown) are: The corresponding calibration parameter mean is (μ ope， μ rea , μ pre , μ m , μ pub ), the standard deviation of the calibration parameter is (σ ope， σ rea , σ pre , σ m , σ pub ), using five kinds of observed value data, namely processing time, standby time, pretreatment time, product quality, and public system power consumption, to calibrate these process parameters with unknown distribution parameters. There are 7 kinds of determined values and 4 kinds of process parameters with accidental uncertainty (distribution parameters are known).
[0067] Can be collected on-site Figure 3 The extrusion molding process parameters in the Bayesian network shown The statistical results of the observed data are shown in Table 1. Based on this, a calibration data set containing 100 observed data is created, and then Markov Monte Carlo sampling (Metropolise-Hastings algorithm) is used to generate samples of the posterior distribution of process parameters.
[0068] Table 1 Statistics of parameters used to generate the calibration dataset
[0069]
[0070] Process parameters The measurement error is assumed to be normally distributed with zero mean and standard deviation of 0.01.
[0071] The calibration process parameters are provided in Table 2 Assumed prior distribution of , because only the upper and lower bounds are available, the prior distribution is assumed to be uniform.
[0072] Table 2 Prior distribution of calibration parameters
[0073]
[0074] Kernel density estimation and Monte Carlo sampling methods are used for Bayesian forward uncertainty propagation, and the posterior distribution of the total energy consumption (E) of the extrusion molding manufacturing system is estimated based on the posterior distribution samples of the process parameters obtained in the Bayesian reasoning part. According to the calibration process parameters provided in Table 2 The same method is used to estimate the prior distribution of the total energy consumption (E) of the extrusion molding manufacturing system. Figure 4 The prior distribution and posterior distribution of the total energy consumption (E) of the manufacturing system are shown in Figure 2. In addition, it can be seen that there is no significant difference between the prior distribution and the posterior distribution, which means that the main source of uncertainty in the energy consumption of the extrusion molding manufacturing system is accidental and therefore irreducible.
[0075] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A manufacturing system energy consumption evaluation method based on Bayesian network, characterized in that: The following steps are involved: Step 1: Obtain the aluminum profile extrusion manufacturing process, which includes a heating furnace, an extruder, processing equipment and a conveying device; Step 2: Obtain the total energy consumption E in the manufacturing process m , the total energy consumption E m Including processing energy consumption E ope , Standby energy consumption E rea , Pretreatment energy consumption E pre , Transportation energy consumption E tran and public energy consumption E pub ; Step 3: Build an evaluation model based on manufacturing process and total energy consumption; Step 4: Obtain the sources of uncertainty and build a prediction model based on accidental uncertainty; Step 5: construct a process Bayesian network based on the multidimensional prediction model, and generate samples of the posterior distribution of process parameters based on the process Bayesian network; Step 6: Use kernel density estimation and Monte Carlo sampling methods to perform Bayesian forward uncertainty propagation, and obtain the posterior distribution of process parameters to estimate the posterior distribution of the total energy consumption of the extrusion molding manufacturing system.
2. The manufacturing system energy consumption evaluation method based on Bayesian network according to claim 1 is characterized in that: The manufacturing process in step 1 is as follows: heating of aluminum bars → extrusion → cooling and quenching → stretching and straightening → sawing → inspection → artificial aging → inspection → warehousing.
3. The manufacturing system energy consumption evaluation method based on Bayesian network according to claim 2 is characterized in that: The processing energy consumption E in step 2 ope is the energy consumption of the extruder and the three subsequent processes when machining aluminum profile product i in order e, 4. The manufacturing system energy consumption evaluation method based on Bayesian network according to claim 3 is characterized in that: The energy consumption of transportation in step 2 is E tran The energy consumed when the product is transferred between the various devices in the extrusion manufacturing system.
5. The manufacturing system energy consumption assessment method based on Bayesian network according to claim 4 is characterized in that: The public energy consumption E in step 2 pub It is the sum of energy consumed by lighting, cooling, water supply and air supply to maintain a constant temperature in the production environment.
6. The manufacturing system energy consumption assessment method based on Bayesian network according to claim 5 is characterized in that: In the step 4, the Bayesian network technology expresses the multivariate joint probability density function as the product of the marginal and conditional probability density functions, which is used for information fusion and the integration of multidimensional uncertainty factors.
7. The manufacturing system energy consumption assessment method based on Bayesian network according to claim 6 is characterized in that: The measurement error of the process parameters in step 6 is assumed to be normally distributed with a zero mean and a standard deviation of 0.01.